fix : validate models before loading

src/tserve/cli/main.py

@@ -121,11 +121,6 @@ def main(argv: list[str] | None = None) -> int:
     ------
     SystemExit
         From ``ArgumentParser.error`` or argparse usage errors.
-    ValueError
-        If positional tokens are malformed (empty ``id=spec``,
-        a bare craft spec), or ``Server`` / ``bootstrap`` reject a
-        load spec (duplicate id, unknown registry id, non-zip path,
-        unknown executor).
     TypeError
         If a model object is not a sktime ``BaseForecaster``.
     ImportError
@@ -150,13 +145,16 @@ def main(argv: list[str] | None = None) -> int:
 
     from tserve.server import Server
 
-    server = Server(
-        model=parse_model(args.positional_model),
-        models_dir=args.models_dir,
-        host=args.host,
-        port=args.port,
-        log_level=args.log_level,
-    )
+    try:
+        server = Server(
+            model=parse_model(args.positional_model),
+            models_dir=args.models_dir,
+            host=args.host,
+            port=args.port,
+            log_level=args.log_level,
+        )
+    except ValueError as exc:
+        parser.error(str(exc))
     try:
         server.run()
     except KeyboardInterrupt:

src/tserve/cli/tests/test_main.py

@@ -107,3 +107,18 @@ def test_main_version(capsys):
 
     assert excinfo.value.code == 0
     assert capsys.readouterr().out.strip() == f"tserve {__version__}"
+
+
+def test_main_reports_invalid_model_without_traceback(capsys):
+    with (
+        patch(
+            "tserve.server.Server", side_effect=ValueError("unknown model 'missing'")
+        ),
+        pytest.raises(SystemExit) as excinfo,
+    ):
+        main(["missing"])
+
+    assert excinfo.value.code == 2
+    err = capsys.readouterr().err
+    assert "unknown model 'missing'" in err
+    assert "Traceback" not in err

src/tserve/runtime/bootstrap.py

@@ -89,11 +89,10 @@ def loaded_models(self) -> ModelsResult:
 def bootstrap(model: list[str | Path | tuple[str, Any]]) -> Runtime:
     """Resolve, construct, load, warmup, and register each selected model.
 
-    For every item: ``resolve_model`` → ``create_executor(info.executor)``
-    → ``load`` → ``warmup`` → ``stats.register``. Tuple items pass the
-    second element (craft spec or object) to ``load``; strings and
-    paths pass the item itself. Duplicate ``ModelInfo.id`` values raise
-    before a second load.
+    Resolve and check all items before creating any executor. Then, for
+    every item: ``create_executor(info.executor)`` → ``load`` → ``warmup``
+    → ``stats.register``. Tuple items pass the second element (craft spec
+    or object) to ``load``; strings and paths pass the item itself.
 
     Parameters
     ----------
@@ -132,6 +131,15 @@ def bootstrap(model: list[str | Path | tuple[str, Any]]) -> Runtime:
     tserve.scheduling.scheduler.Scheduler
         Forecast dispatch is not implemented in this package.
     """
+    resolved: list[tuple[ModelInfo, Any]] = []
+    seen: set[str] = set()
+    for item in model:
+        info = resolve_model(item)
+        if info.id in seen:
+            raise ValueError(f"duplicate model id {info.id!r} in model")
+        seen.add(info.id)
+        resolved.append((info, item[1] if isinstance(item, tuple) else item))
+
     stats = Stats()
     executors: dict[str, Executor] = {}
     models: dict[str, ModelInfo] = {}
@@ -145,13 +153,7 @@ def bootstrap(model: list[str | Path | tuple[str, Any]]) -> Runtime:
             f"{paint('is always loaded as a baseline', '2')}"
         )
 
-    for position, item in enumerate(model, start=1):
-        info = resolve_model(item)
-        item = item[1] if isinstance(item, tuple) else item
-
-        if info.id in models:
-            raise ValueError(f"duplicate model id {info.id!r} in model")
-
+    for position, (info, item) in enumerate(resolved, start=1):
         label = f"{info.id} via {info.executor}"
         dots = paint("." * max(3, 40 - len(label)), "2")
         prefix = (

src/tserve/runtime/tests/test_bootstrap.py

@@ -77,10 +77,31 @@ def test_bootstrap_loads_craft():
 def test_bootstrap_rejects_duplicate():
     with (
         patch("tserve.runtime.bootstrap.resolve_model", return_value=_info()),
-        patch("tserve.runtime.bootstrap.create_executor", return_value=_executor()),
+        patch("tserve.runtime.bootstrap.create_executor") as create_executor,
         pytest.raises(ValueError, match="duplicate model id 'naive' in model"),
     ):
         bootstrap(["naive", "naive"])
+    create_executor.assert_not_called()
+
+
+def test_bootstrap_validates_before_loading():
+    executor = _executor()
+    with (
+        patch(
+            "tserve.runtime.bootstrap.resolve_model",
+            side_effect=[_info(), ValueError("unknown model 'missing'")],
+        ) as resolve_model,
+        patch(
+            "tserve.runtime.bootstrap.create_executor", return_value=executor
+        ) as create_executor,
+        pytest.raises(ValueError, match="unknown model 'missing'"),
+    ):
+        bootstrap(["naive", "missing"])
+
+    assert resolve_model.call_count == 2
+    create_executor.assert_not_called()
+    executor.load.assert_not_called()
+    executor.warmup.assert_not_called()
 
 
 def test_loaded_models():